Introduction

Meat inspection plays a key role in Belgium’s animal disease surveillance and public health system (García-Díez et al. 2023). Its main goals are to protect human health by identifying animals unfit for consumption, removing them from the food chain, and monitoring animal diseases. The effectiveness of this system heavily relies on the sensitivity (Se) of post-mortem inspections (Welby et al. 2012).

However, Se has not been directly evaluated at the slaughterhouse level in Belgium. For bovine tuberculosis—a notifiable disease—modeling studies suggest low sensitivity: between 2010 and 2015, only 16 suspicious lesions were reported annually on average, despite estimates ranging from 26 to 4,684 expected cases per year (Welby et al. 2022). This raises concerns about the ability of meat inspection to detect not only notifiable but also non-notifiable diseases, where Se may be even lower.

There is also substantial variability in detection rates between slaughterhouses, likely influenced by factors such as staff experience (Pozo et al. 2021). Furthermore, certain exotic or early-stage diseases, or pathogens without visible lesions, may go undetected through conventional inspection methods (Stärk et al. 2014). Given that Belgium relies heavily on slaughterhouse surveillance, this poses a risk to early detection and disease control.

This project aims to evaluate the performance of cattle disease surveillance at the slaughterhouse level in Belgium, focusing on lesion detection. The specific objectives are:

  • Characterization of Belgian slaughterhouses (Work Package 1)

  • Improving lesion reporting systems (Work Package 2)

  • Identifying risk factors associated with lesion detection (Work Package 3)

  • Evaluating the sensitivity of lesion detection per slaughterhouse using bovine tuberculosis as a case study (Work Package 4)

Data

This study leverages the integration and matching of data from three key sources: BELTRACE, NEOLIMS, and SANITEL, covering the period from January 1, 2019, to October 21, 2024 (date of the last data extraction). BELTRACE database provides detailed information on slaughtered cattle and the results of post-mortem inspections conducted in slaughterhouses, including detected lesions and seizure decisions. NEOLIMS contains epidemiological data from Sciensano’s veterinary laboratory reports, such as animal disease screening test results. SANITEL complements these with administrative and traceability data, including herd identification and the movement history of slaughtered animals. In addition, data were collected on the work of slaughterhouse inspectors through an online questionnaire. Finally, some additional data concerning the time and place of work of the inspectors has been provided by the Federal Agency for the Safety of the Food Chain (FASFC).

WP1 Descriptive and trend analysis of slaughtered cattle in belgium

Summary dataset

Average monthly slaughtered cattle per slaughterhouse in Belgium, 2019-2024

Cattle slaughtering per month in Belgium from January 2019 to September 2024

Age distribution of slaughtered cattle in Belgium

Origin of cattle slaughtered in belgium

Over the past five years, 79.96% of cattle slaughtered in Belgium originated from Belgian farms. The remaining 20% predominantly came from the Netherlands, as illustrated in Figure below. Notably, the vast majority of cattle imported from the Netherlands are calves.

Origin of cattle slaughtered in belgium

WP2 : SWOT Analysis

A questionnaire was developed for post-mortem meat inspectors.
A total of 60 people clicked on the link to access the questionnaire. Five stopped after reading the project introduction, and five opened the questionnaire but did not submit any responses. In fine, we received and analysed 39 complete questionnaires et 11 incomplete ones (i.e. at least one answer). One incomplete questionnaire was excluded as the respondent works in a poultry slaughterhouse and only completed Section A (general information). All remaining questionnaires (complete or not) were analysed.
The questionnaire results were analyzed and contributed to the SWOT analysis presented below.

WP3 : Risk sur détection de lésion PM

An univariate logistic regression was performed on all those potential risk factors to determine whether they have a potential statistically significant effect on post-mortem lesion detection at the slaughterhouse. Logistic regression is commonly used in statistical modelling because it represents the relationship between a set of explanatory variables (e.g., age, sex, etc.) and a binary dependent variable (in our case anomaly detection Yes/No).

Recap parameter

The results of each univariate analysis for the various discrete parameters considered in this project are presented below.
In this table, variables in red characters are reference values.

Slaughter line speed

Parameter of the logistic regression and theorical curve

(Intercept) vitesse_chaine
0.1576 0.9788

Plot of Estimated Anomaly Detection Probabilities and Observed Data Proportions by Decile for Slaughter Line Speed

Age of the slaughtered cattle

Parameter of the logistic regression and theorical curve

(Intercept) age_abattu_mois
0.06201 1.012

Plot of Estimated Anomaly Detection Probabilities and Observed Data Proportions by Age Decile of Slaughtered Cattle

Slaughterhouse

Estimated probabilities per Slaughterhouse

WP4 Sensitivity Slaughterhouse for bTB

In Scenario 1, a fixed sensitivity of 40% was assumed. However, younger animals tend to have less developed lesions, making detection more difficult compared to older animals. To address this, Scenario 2 applied an age-based linear sensitivity model, ranging from 20% for animals under 1 year old to 70% for those aged 4 years and above. These values were informed by expert opinion and the sensitivity range (0.38–0.92) reported by Welby et al. (2022). Figure below illustrates the theoretical sensitivity curve used in Scenario 2, showing how detection improves with increasing animal age.

sensitivity curve used in Scenario 2

Quality of detection in the 43 evaluated cattle slaughterhouses :

  • 8 slaughterhouses (18.6%) in Scenario 1 (where sensitivity was fixed at 40%) but also in Scenario 2 (the same slaughterhouses) have a ratio equal to or greater than 1. These slaughterhouses are highlighted in green in Table below. They can be considered as reporting enough suspicions.

  • 16 slaughterhouses had a ratio < 1, meaning that the number of reported suspicion was below the expected number (highlighted in yellow).

  • 4 slaughterhouses should be flagged, in our view, as concerning, due to under-reporting (highlighted in pink in Table below) : three of them did not report any suspicion over the five-year period despite being expected to report at least eight. The fourth slaughterhouse is particularly concerning, as it reported only one suspicion, whereas it was expected to report at least 164 in Scenario 1 or 141 in Scenario 2.

  • For the remaining 15 slaughterhouses (lines in italic), the total number of expected suspicions was <1 (values in red characters in Table III), hence it was logical that these slaughterhouses did not report any suspicion, due to a low number of slaughtered animals (ratio not calculated = NC).

Conclusion

The detection of animal diseases post-mortem at the slaughterhouse is a key component of animal disease surveillance and public health in Belgium. However, certain factors that remain insufficiently addressed today negatively impact the inspection of cattle carcasses. With more than 4.5 million cattle slaughtered between 2019 and 2024 (nearly 80% of which originated from Belgium), this project revealed significant disparities in the age and profiles of slaughtered cattle across different slaughterhouses. Over this six-year period, no consistent trend emerged from the monthly slaughter figures, apart from recurrent fluctuations in certain months, likely linked to the number of working days.

Data analysis highlights substantial variability in the detection of suspected lesions during post-mortem investigation, with reporting rates ranging from 0.0025% to 29.04% across slaughterhouses. This disparity appears to be driven by several key factors: working conditions, inspector training and experience, the tools available during inspections, and the characteristics of the cattle processed. Considering bTB, our analysis revealed considerable heterogeneity in the reporting of suspicions between slaughterhouses. A small number of slaughterhouses consistently met or exceeded expectations based on model assumptions, while several others reported little to no suspicions, suggesting potential under-reporting. This discrepancy highlights areas for improvement in surveillance practices and reinforces the importance of maintaining systematic suspicion reporting, even when confirmation rates remain low.

Survey results emphasize the crucial role of inspector training and experience, which are perceived as major strengths of the current system. However, further targeted training could enhance anomaly detection even more. Interestingly, statistical analyses suggest that as inspectors gain experience, their suspicion rates decrease — a phenomenon whose underlying causes remain to be clarified. Working conditions — including space, lighting, and especially slaughter line speed — emerge as major areas for improvement. A high processing speed significantly reduces inspectors’ ability to detect lesions, underscoring the need to optimize these conditions to safeguard vigilance and accuracy. While the introduction of specialized inspection tools could ease physical strain and help offset high processing speeds, their use in Belgium remains limited.

Technical challenges in database management also became apparent during this project. The lack of harmonization and standardized encoding protocols complicates data integration, leading to loss of information, inaccuracies, and limited analytical possibilities. Improved data structuring and interoperability are essential to enhance surveillance reliability and decision-making capacity.

To optimize post-mortem lesion detection in slaughterhouses, it is crucial to simultaneously address several aspects: improving inspectors’ working conditions, strengthening continuous professional development, and developing appropriate tools. At the same time, modernizing and standardizing data management systems would significantly improve the accuracy of analyses and reinforce the effectiveness of animal health surveillance and public health protection in Belgium.

References

García-Díez, Juan, Sónia Saraiva, Dina Moura, Luca Grispoldi, Beniamino Terzo Cenci-Goga, and Cristina Saraiva. 2023. “The Importance of the Slaughterhouse in Surveilling Animal and Public Health: A Systematic Review.” Veterinary Sciences 10 (2): 167. https://doi.org/10.3390/vetsci10020167.

Pozo, Pilar, Nicolas Cespedes Cardenas, Javier Bezos, Beatriz Romero, Anna Grau, Jesus nacar, Jose Luis Saez, Olga Minguez, and Julio Alvarez. 2021. “Evaluation of the Performance of Slaughterhouse Surveillance for Bovine Tuberculosis Detection in Castilla y Leon, Spain.” Preventive Veterinary Medicine 189 (April):105307. https://doi.org/10.1016/j.prevetmed.2021.105307.

Stärk, K.D.C., S. Alonso, N. Dadios, C. Dupuy, L. Ellerbroek, M. Georgiev, J. Hardstaff, et al. 2014. “Strengths and Weaknesses of Meat Inspection as a Contribution to Animal Health and Welfare Surveillance.” Food Control 39 (May):154–62. https://doi.org/10.1016/j.foodcont.2013.11.009.

Welby, S., M. Govaerts, L. Vanholme, J. Hooyberghs, K. Mennens, L. Maes, and Y. Van Der Stede. 2012. “Bovine Tuberculosis Surveillance Alternatives in Belgium.” Preventive Veterinary Medicine 106 (2): 152–61. https://doi.org/10.1016/j.prevetmed.2012.02.010.

Welby, Sarah, Mickaël Cargnel, and Claude Saegerman. 2022. “Quantitative Decision Making in Animal Health Surveillance: Bovine Tuberculosis Surveillance in Belgium as Case Study.” Transboundary and Emerging Diseases 69 (4). https://doi.org/10.1111/tbed.14269.